Share
Embed
Lecture №2.1. Templates.
1. Motivation for templates.
2. How it work? Instantiation.
3. Template argument deduction from a function call.
5. Class templates.
6. Function templates specialization.
7. Class templates specialization.
8. Member templates.
9. Traits classes.
10. Policy classes.
11. Functors from functional. std::bind.
12. Lambda-expression.
13. SFINAE.
14. Iterator implementation until and since C++17.
Baramiya Denis
September 03, 2019
Transcript
51 slides
-
LANGUAGE C #define max(x,y) ((x) > (y) ? (x) :
(y)) #define min(x,y) ((x) < (y) ? (x) : (y)) -
LANGUAGE C++ inline int max(int x, int y) { return
x > y ? x : y; } inline long max(long x, long y) { return x > y ? x : y; } inline float max(float x, float y) { return x > y ? x : y; } inline double max(double x, double y) { return x > y ? x : y; } // ......??????? -
FUNCTION TEMPLATES template <parameter-list> function-declaration template <typename T> inline T
const &max(T const &x, T const &y) { return x > y ? x : y; } template <class T> inline T const &min(T const &x, T const &y) { return x < y ? x : y; } -
FUNCTION TEMPLATES template <parameter-list> function-declaration template <typename T> inline T
const &max(T const &x, T const &y) { return x > y ? x : y; } template <class T> inline T const &min(T const &x, T const &y) { return x < y ? x : y; } typename and class are equivalent. -
int a = 10; int b = min(a, 7); template
<typename T> inline T const &min(T const &x, T const &y) { return x < y ? x : y; } A compiler is looking for min function. Finds this template and infers type T by type of function arguments. -
int a = 10; int b = min(a, 7); inline
int const &min(int const &x, int const &y) { return x < y ? x : y; } template <typename T> inline T const &min(T const &x, T const &y) { return x < y ? x : y; } A compiler is looking for min function. Finds this template and infers type T by type of function arguments. Template instantiation -
int a = 10; int b = min(a, 7); inline
int const &min(int const &x, int const &y) { return x < y ? x : y; } template <typename T> inline T const &min(T const &x, T const &y) { return x < y ? x : y; } A compiler is looking for min function. Finds this template and infers type T by type of function arguments. Template instantiation Instance Function Substitution -
min(1, 1.2); // Error: different types min(double(1), 1.2); // OK.
typecasting min<double>(1, 1.2); // OK. explicit type T -
MULTIPLE PARAMETERS template <typename T1, typename T2> inline T1 const
&min(T1 const &x, T2 const &y) { return x < y ? x : y; } // .... min(1, 1.2); // OK. The return type is determined // by the type of x. template <typename RT, typename T1, typename T2> inline RT const &max(T1 const &x, T2 const &y) { return x > y ? x : y; } // .... max<double>(4, 4.2); // Type inference is not possible for RT -
inline int const &max(int const &x, int const &y) {
return x > y ? x : y; } template <typename T> inline T const &max(T const &x, T const &y) { return x > y ? x : y; } template <typename T> inline T const &max(T const &x, T const &y, T const &z) { return max(max(x,y), z); } int main() { max(7, 42, 68); // use template with 3 arguments max(7.0, 42.0); // max<double> (template argument deduction) max('a', 'b'); // max<char> (template argument deduction) max(7, 42); // usual function max<>(7, 42); // max<int> (template argument deduction) max<double>(7, 42); // max<double> max('a', 42.7); // usual function with 2 int arguments } FUNCTION TEMPLATE OVERLOADING -
// From <algorithm> template<class InpIt, class OutIt> OutIt copy(InpIt first,
InpIt last, OutIt result) { while (first != last) { *result = *first; ++result; ++first; } return result; } -
CLASS TEMPLATES template <typename T> class Stack { std::vector<T> elems;
public: void push(T const &); void pop(); T top() const; bool empty() const { return elems.empty() } }; template <typename T> void Stack<T>::push(T const &elem) { elems.push_back(elem); } template <typename T> void Stack<T>::pop() { if (elems.empty()) throw std::out_of_range("empty stack"); elems.pop_back(); } template <typename T> T Stack<T>::top() const { if (elems.empty()) throw std::out_of_range("empty stack"); elems.back(); } -
int main() { try { Stack<int> intStack; Stack<std::string> stringStack; intStack.push(7);
stringStack.push("hello"); stringStack.pop(); stringStack.pop(); } catch (std::exception const &ex) { std::cerr << "Exception " << ex.what() << std::endl; return -1; } } -
FUNCTION TEMPLATE SPECIALIZATION template<typename T> inline void exchange(T *a, T
*b) { T tmp(*a); *a = *b; *b = tmp; } // .... void swap_arrays(Array<int> *a1, Array<int> *a2) { exchange(a1, a2); } template<typename T> class Array { T *data; public: // ... void exchange_with(Array<T> *b) { T *tmp = data; data = b->data; b->data = tmp; } }; Как заставить exchange использовать Array::exchange_with? -
// template 1 template<typename T> inline void quick_exchange(T *a, T
*b) { T tmp(*a); *a = *b; *b = tmp; } // template 2 template<typename T> inline void quick_exchange(Array<T> *a, Array<T> *b) { a->exchange_with(b); } void demo(Array<int> *a1, Array<int> *a2) { int x = 1, y = 2; quick_exchange(&x, &y); // template 1 quick_exchange(a1, a2); // template 2 } «More specialized template» for a C++ compiler -
CLASS TEMPLATE SPECIALIZATION template <typename T> class Storage8 { T
objects[8]; public: void set(int idx, const T &t) { objects[idx] = t; } const T& operator[](int idx) { return objects[idx]; } }; Storage8<bool> can be optimized. -
template <> class Storage8<bool> { unsigned char bits; public: void
set(int idx, bool t) { unsigned char mask = 1 << idx; if (t) bits |= mask; else bits &= ~mask; } bool operator[](int idx) { return bits & (1 << idx); } }; -
PARTIAL TEMPLATE SPECIALIZATION • Класс List будет инстанцироваться для всех
вариантов T. В большом проекте это может привести к разбуханию кода (code bloat). • С низкоуровневой точки зрения реализации List<int *>::append() и List<void *>::append() идентичны. • Нельзя ли использовать этот факт для оптимизации списков указателей? template <typename T> class List { public: // ... void append(T const &); inline size_t length() const; // ... }; -
// Explicit or full specialization of List<void *> template<> class
List<void *> { // ... void append(void *p); inline size_t length() const; // ... }; // Partial specialization of List<T *> template <typename T> class List<T *> { List<void *> impl; public: // ... void append(T *p) { impl.append(p); } inline size_t length() const { return impl.length(); } // ... }; -
template <typename T> class Vector { T *base; public: //
... // Dynamic polymorphism void print(ostream &os) { for (auto const &v : *this) os << v; } // Static polymorphism template <typename Out> void print(Out &out) { for (auto const &v : *this) out << v; } }; MEMBER TEMPLATES -
TRAITS CLASSES • Шаблоны позволяют иметь произвольное количество аргументов. •
Можно настроить любой аспект поведения класса или функции. • Но передавать всегда и везде по 100 аргументов неудобно… -
WAT template <typename T> inline T accum(const T *beg, const
T *end) { T total = T(); // T() возвращает ноль // Но лучше писать T{} while (beg != end) total += *beg++; return total; } /////// int num[] = { 1, 2, 3, 4, 5 }; int avg1 = accum(&num[0], &num[5]) / 5; // 3 char name[] = "templates"; int len = sizeof(name) - 1; int avg2 = accum(&name[0], &name[len]) / len; // -5 -
• Переменная total имеет тип char (8 бит), которого не
хватает для суммирования. • Можно тип суммы сделать аргументом шаблона… но это очень неудобно. • Воспользуемся другим подходом. -
template <typename T> class AccumulationTraits; template<> class AccumulationTraits<char> { public:
typedef int AccT; }; template<> class AccumulationTraits<short> { public: typedef int AccT; }; template<> class AccumulationTraits<int> { public: typedef long AccT; }; // unsigned int -> unsigned long // float -> double // ..... -
template <typename T> inline typename AccumulationTraits<T>::AccT accum(const T *beg, const
T *end) { typedef typename AccumulationTraits<T>::AccT AccT; AccT total = AccT(); // AccT() возвращает нулевое значение while (beg != end) total += *beg++; return total; } char name[] = "templates"; int len = sizeof(name) - 1; int avg2 = accum(&name[0], &name[length]) / len; // 108 -
template <typename T> class AccumulationTraits; template<> class AccumulationTraits<char> { public:
typedef int AccT; static AccT zero() { return 0; } }; template<> class AccumulationTraits<short> { public: typedef int AccT; static AccT zero() { return 0; } }; template<> class AccumulationTraits<int> { public: typedef long AccT; static AccT zero() { return 0; } }; template <typename T> inline typename AccumulationTraits<T>::AccT accum(const T *beg, const T *end) { typedef typename AccumulationTraits<T>::AccT AccT; AccT total = AccumulationTraits<T>::zero(); while (beg != end) total += *beg++; return total; } Member function zero() -
template <typename T, typename AT = AccumulationTraits<T> > inline typename
AT::AccT accum(const T *beg, const T *end) { // ... } Traits as template parameters -
• Вместо того чтобы свойства (типы, константы, …) делать параметрами
шаблона по T, помещаем их в дополнительный класс свойств (Traits). • Создаем набор полных специализаций Traits для всех нужных T. TRAITS CLASSES -
#include <type_traits> #include <iostream> #include <array> #include <string> #include <type_traits>
using namespace std; int main() { cout << boolalpha; cout << "is_array:" << endl; cout << "int: " << std::is_array<int>::value << endl; cout << "int[3]: " << std::is_array<int[3]>::value << endl; cout << "array<int,3>: " << std::is_array<array<int,3>>::value << endl; cout << "string: " << std::is_array<string>::value << endl; cout << "string[3]: " << std::is_array<string[3]>::value << endl; return 0; } is_array: int: false int[3]: true array<int,3>: false string: false string[3]: true -
POLICY CLASSES Операция суммирования вынесена в Policy class SumPolicy {
public: template <typename T1, typename T2> static void accumulate(T1 &total, const T2 &value) { total += value; } }; template <typename T, typename Policy = SumPolicy, typename Traits = AccumulationTraits<T> > inline typename Traits::AccT accum(const T *beg, const T *end) { typename Traits::AccT total = Traits::zero(); while (beg != end) { Policy::accumulate(total, *beg); ++beg; } return total; } -
template <bool use_compound_op = true> class SumPolicy { public: template
<typename T1, typename T2> static void accumulate(T1 &total, const T2 &value) { total += value; } }; template <> class SumPolicy<false> { public: template <typename T1, typename T2> static void accumulate(T1 &total, const T2 &value) { total = total + value; } }; class MultPolicy { public: template <typename T1, typename T2> static void accumulate(T1 &total, const T2 &value) { total *= value; } }; Умножение Тонкая настройка суммирования -
#include <iostream> // std::cout int main() { int first[] =
{ 1, 2, 3, 4, 5}; int second[] = { 10, 20, 30, 40, 50}; int results[5]; for (int i = 0; i < 5; ++i) results[i] = first[i] + second[i]; for (int i = 0; i < 5; i++) std::cout << results[i] << ' '; std::cout << '\n'; return 0; } -
#include <iostream> // std::cout #include <functional> // std::plus #include <algorithm>
// std::transform int main() { int first[] = { 1, 2, 3, 4, 5}; int second[] = { 10, 20, 30, 40, 50}; int results[5]; std::transform(first, first+5, second, results, std::plus<int>()); for (int i = 0; i < 5; i++) std::cout << results[i] << ' '; std::cout << '\n'; return 0; } How fix? -
#include <iostream> // std::cout #include <functional> // std::plus #include <algorithm>
// std::transform int main() { int first[] = { 1, 2, 3, 4, 5}; int second[] = { 10, 20, 30, 40, 50}; int results[5]; std::transform(first, first+5, second, results, std::plus<int>()); std::copy(result, result+5, std::ostream_iterator<int>(std::cout, ' '); std::cout << '\n'; return 0; } -
template <class _Arg1, class _Arg2, class _Result> struct binary_function {
typedef _Arg1 first_argument_type; ///< the type of the first argument /// (no surprises here) typedef _Arg2 second_argument_type; ///< the type of the second argument typedef _Result result_type; ///< type of the return type }; template <class _Tp> struct plus : public binary_function<_Tp, _Tp, _Tp> { _Tp operator()(const _Tp& __x, const _Tp& __y) const { return __x + __y; } }; -
#include <iostream> #include <functional> int main() { int ary[] =
{ 1, 2, 3, 4, 5 }, res[5]; using namespace std::placeholders; // _1, _2, _3,... auto inc_10 = std::bind(std::plus<int>(), _1, 10); std::transform(ary, ary+5, res, inc_10); for (int x: res) std::cout << x << std::endl; // 11… 12… 13… 14… 15 return 0; } } Currying -
bool all_under_20 = std::all_of(ary, ary + 5, [](int n) {
return n < 20; }); Lambda-expression -
// Сколько элементов вектора v принадлежат отрезку [loBo, upBo)? size_t
rangeMatch(const vector<int> &v, int loBo, int upBo) { return std::count_if(v.begin(), v.end(), [loBo, upBo](int _n) { return loBo <= _n && _n < upBo; }); } "Capturing" variables -
[loBo, upBo](int _n) { return loBo <= _n && _n
< upBo; } // примерно соответствует: struct AutomaticallyGenerated { AutomaticallyGenerated(int lo, int up) : loBo(lo), upBo(up) {} bool operator()(int _n) { return loBo <= _n && _n < upBo; } int loBo, upBo; }; -
#include <iostream> #include <functional> int main() { int ary[] =
{ 1, 2, 3, 4, 5 }, res[5]; auto incGen = [] (int _val) -> std::function<int (int)> { return [_val] (int _n) -> int { return _n + _val; }; }; auto inc_10 = incGen(10); std::transform(ary, ary+5, res, inc_10); for (int x: res) std::cout << x << std::endl; // 11… 12… 13… 14… 15 return 0; } } Lambda-expression generates lambda- expressions -
SFINAE Substitution Failure Is Not An Error struct Bar{ typedef
double internalType; }; template <typename T> typename T::internalType foo(const T& t) { std::cout << "foo<T>" << std::endl; return 0; } int main(){ foo(Bar()); foo(0); // error // no matching function for call to 'foo(int)' // ... // template argument deduction/substitution failed: } -
ENABLE_IF std::enable_if<condition, T>::type template <typename T> typename std::enable_if<std::is_arithmetic<T>::value, T>::type foo(T
t) { std::cout << "foo<arithmetic T>" << std::endl; return t; } -
ITERATORS // Until С++17 class num_iterator : std::iterator<std::forward_iterator_tag, int> {
int i; public: explicit num_iterator(int pos = 0) : i{ pos } {} int operator*() const { return i; } num_iterator& operator++() { ++i; return *this; } bool operator!=(const num_iterator &other) const { return i != other.i; } bool operator==(const num_iterator &other) const { return !(*this != other); } }; // Since С++17 std::iterator - deprecated class num_iterator { int i; public: explicit num_iterator(int pos = 0) : i{ pos } {} ... }; namespace std { template <> struct iterator_traits<num_iterator> { using iterator_category = std::forward_iterator_tag; using value_type = int; //using pointer = ...; //using reference = ...; //using difference_type = ...; }; }